All Discussions > Steam Forums > Hardware and Operating Systems > Topic Details
r.linder Jul 21, 2019 @ 5:26pm
3DMark (Time Spy) - Someone PLEASE Explain This
https://www.3dmark.com/compare/spy/7868685/spy/5181165#

The lower score build that was taken today is from my secondary system. I left the 2700X at stock and enabled PBO as I have no spare coolers that can actually handle a manual overclock for a Ryzen chip, but I overclocked my 1070 Ti with MSI afterburner to as high as it could go without raising voltage. You can see from the comparison that my 1070 Ti has much higher clocks versus the top score for this specific config (2700X and 1 1070 Ti), as well as a higher boost clock, but scores are poorer across the board.

I can't wrap my head around it. I get that the CPU boosts are on a single core with 1st and 2nd gen Ryzen, but that doesn't explain how a 1070 Ti clocked way lower completely destroys it in the same benchmark. I've seen no such rift when it came to driver versions in my main system (3900X, 2080, being in the top ten scores for all 3DMark tests so far), so I doubt it's related to drivers, unless NVIDIA completely ♥♥♥♥ed over Pascal.

2700X
Gigabyte B450M-DS3H
16GB G.Skill Trident Z RGB DDR4-3000 CL15
MSI GTX 1070 Ti Titanium (+200 core, +550 memory)
ADATA Ultimate SU800 512GB SATA SSD
Last edited by r.linder; Jul 21, 2019 @ 5:48pm
< >
Showing 1-15 of 18 comments
emoticorpse Jul 21, 2019 @ 5:38pm 
Motherboard/SSD difference?.
#1
r.linder Jul 21, 2019 @ 5:44pm 
Originally posted by emoticorpse:
Motherboard/SSD difference?.

B450M-DS3H vs PRIME X470-PRO
I know the motherboard doesn't affect the result, I had the same score range before on an X470 board before RTX released, the exact same GPU and CPU that was in those tests.

512GB ADATA Ultimate SU800 SATA SSD vs 250GB Samsung 960 EVO
SSD shouldn't create a 4000 point difference. There's no benefit for this above a SATA SSD.
Last edited by r.linder; Jul 21, 2019 @ 5:45pm
#2
hawkeye Jul 21, 2019 @ 5:47pm 
User error - using different hardware and software and expecting same result ?

Reset it to default. Run one test, overclock it, compare result.
Last edited by hawkeye; Jul 21, 2019 @ 5:48pm
#3
r.linder Jul 21, 2019 @ 5:49pm 
Originally posted by hawkeye:
User error - using different hardware and software and expecting same result ?

User error? Did you even look at the comparison? My hardware is clocked higher.
#4
hawkeye Jul 21, 2019 @ 5:54pm 
Originally posted by 【Ｅｓｃｏｒｖｅ】:
Originally posted by hawkeye:
User error - using different hardware and software and expecting same result ?

User error? Did you even look at the comparison? My hardware is clocked higher.

Benchmarking is about taking a baseline and then changing the one factor that you want to test.

You have different windows versions, different hardware and you haven't even supplied any info about what and how you changed it. And you want us to gaze into a crystal ball? If you aren't going to do it scientifically, what can you expect?
Last edited by hawkeye; Jul 21, 2019 @ 5:58pm
#5
r.linder Jul 21, 2019 @ 6:05pm 
Originally posted by hawkeye:
Originally posted by 【Ｅｓｃｏｒｖｅ】:

User error? Did you even look at the comparison? My hardware is clocked higher.

Benchmarking is about taking a baseline and then changing the one factor that you want to test.

You have different windows versions, different hardware and you haven't even supplied any info about what and how you changed it. And you want us to gaze into a crystal ball? If you aren't going to do it scientifically, what can you expect?

I already stated it was the same hardware configuration. Both tests included the same CPU and GPU, with only difference in motherboard, SSD quality, and 200 MHz in RAM speed, which I know doesn't make a 4000 point difference because I tested multiple kits of RAM with Ryzen. I used the exact same chip and GPU on a higher quality motherboard and SSD and got basically the same result, as well as with CL14 B-die memory. There is no "scientific" reason why hardware clocked considerably lower would be getting such a high score.
#6
Spec_Ops_Ape Jul 21, 2019 @ 6:12pm 
Do your clock speeds actually stick during the run?
Whats the power doing during this time? HWInfo can show this.
#7
r.linder Jul 21, 2019 @ 6:14pm 
Originally posted by Spec_Ops_Ape:
Do your clock speeds actually stick during the run?
Whats the power doing during this time? HWInfo can show this.

Since it's stock/PBO, no. Clocks bounce around and 3DMark is only recording one of the cores. Either way, it's not going to make a 4000 point difference.
Last edited by r.linder; Jul 21, 2019 @ 6:14pm
#8
emoticorpse Jul 21, 2019 @ 6:15pm 
Originally posted by 【Ｅｓｃｏｒｖｅ】:
Originally posted by hawkeye:

Benchmarking is about taking a baseline and then changing the one factor that you want to test.

You have different windows versions, different hardware and you haven't even supplied any info about what and how you changed it. And you want us to gaze into a crystal ball? If you aren't going to do it scientifically, what can you expect?

I already stated it was the same hardware configuration. Both tests included the same CPU and GPU, with only difference in motherboard, SSD quality, and 200 MHz in RAM speed, which I know doesn't make a 4000 point difference because I tested multiple kits of RAM with Ryzen. I used the exact same chip and GPU on a higher quality motherboard and SSD and got basically the same result, as well as with CL14 B-die memory. There is no "scientific" reason why hardware clocked considerably lower would be getting such a high score.

well, from what I can tell most similar hardware setup scores align closer to yours than his, so it isn't so much a question as to why yours is so low. It's why his is so high. My score makes sense compared to yours and nowhere near his and I have a 1700x with 2070 stock clock. I mean there is an explanation as to what he fed his system to get it like that but technically we can only speculate. Maybe Nvidia is pulling an Apple with their software.
Last edited by emoticorpse; Jul 21, 2019 @ 6:18pm
#9
Spec_Ops_Ape Jul 21, 2019 @ 6:59pm 
Your VRAM bus speed is running at more than 16Gbps (2277MHz), the default is 8Gbps (1000MHz per memory die). Might explain why your GPU score suffers so.
#10
r.linder Jul 21, 2019 @ 7:37pm 
Originally posted by Spec_Ops_Ape:
Your VRAM bus speed is running at more than 16Gbps (2277MHz), the default is 8Gbps (1000MHz per memory die). Might explain why your GPU score suffers so.

Any proof of that causing worse performance?
Last edited by r.linder; Jul 21, 2019 @ 7:37pm
#11
Spec_Ops_Ape Jul 21, 2019 @ 7:44pm 
The best you will see on the web is just under 9Gbps (aircooled) for the memory of the 1070 Ti, there is also a point of diminishing returns when overclocking but I assume you already know this.
Last edited by Spec_Ops_Ape; Jul 21, 2019 @ 7:44pm
#12
r.linder Jul 21, 2019 @ 7:49pm 
Originally posted by Spec_Ops_Ape:
The best you will see on the web is just under 9Gbps (aircooled) for the memory of the 1070 Ti, there is also a point of diminishing returns when overclocking but I assume you already know this.

There was a lower score with the default memory clock of 1607. That's how it is literally out of the box.
#13
Spec_Ops_Ape Jul 21, 2019 @ 7:56pm 
You seem intent on making excuses and assumptions to try to fix your perceived problem. As hawkeye said, you need more data to try to pinpoint the issue, not guesswork.
#14
r.linder Jul 21, 2019 @ 8:03pm 
Originally posted by Spec_Ops_Ape:
You seem intent on making excuses and assumptions to try to fix your perceived problem. As hawkeye said, you need more data to try to pinpoint the issue, not guesswork.

I never once asked for a solution, merely an explanation with actual evidence that the idea could be a major source. All you guys haven given is nothing but guesswork, so you're one to talk.
#15
< >
Showing 1-15 of 18 comments
Per page: 1530 50

All Discussions > Steam Forums > Hardware and Operating Systems > Topic Details
Date Posted: Jul 21, 2019 @ 5:26pm
Posts: 18
Start a New Discussion

Discussions Rules and Guidelines